APPLICANT'S ABSTRACT: A recent methodological advance in structural equation modeling is that of latent variable mixture modeling where a population under study is partitioned into an unknown number of latent classes and with which, different means and covariance structures are associated. Integrating methodological works from the mixture modeling literature, Muthen and colleagues (Muthen, in press; Muthen et al., 1998) proposed a general growth mixture modeling framework appropriate for analyzing change in data sets that involve heterogeneous subpopulations. The approach combines latent class and latent growth models and provides a framework for modeling growth functions of varying subpopulations, each with its own growth shape, predictors, concurrent covariates, and outcomes. This newly developed methodology is of practical importance in understanding heterogeneity corresponding to qualitatively different developmental processes. In this proposal, we will make use of the growth mixture modeling method to examine multiple developmental pathways in adolescent alcohol and other substance use and related problem behaviors, using existing longitudinal data sets. Specifically, within the theoretical framework of the social contextual model posited by Patterson et al.(1992), the proposed study will extend current research by examining developmental heterogeneity in adolescent alcohol use and other related problem behaviors (e.g., antisocial behavior) and by identifying varying influences (e.g., parental practices, peer associations, and other common risk and protective factors) associated with heterogeneous populations.